Modeling the growth of Chinese cabbage using remote sensing system

Background: Chinese cabbage of the typical crop in Asia is crucial source of supply containing abundant fibroid materials, minerals. It is required to forecast and diagnose growth by remote sensing (RS), GPS and GIS for precision production management of Chinese cabbage.

The objective of this study was to develop model for estimating the growth (Fresh weight, leaf area) with spectral information of Chinese cabbage in multispectral image acquired by using unmanned aerial system depending on vegetation stages.

Methods: Chinese cabbage in test field was planted on normal planting period (NP) and two delayed planting periods (DP). Multispectral images of test field were acquired at intervals of 2 weeks in midday by using NIR and red edge camera mounted on UAV. Vegetation index and band ratio was calculated by spectral information of sample extracted by image processing. Simple linear regression analysis was employed to develop model of relationship between the growth and index, ratio.

Results: It was developed to two models for accurately estimating the growth because models between NP and DP show different linearity in all vegetation stage. The model using NDVI calculated by NIR and Red bands exhibited lower performance (R2=0.690 and RMSE=868.4g in NP, R2=0.851 and RMSE=285.1g in DP) by saturation of light-sensitivity NIR. The model using ratio between less light-sensitivity Red edge and blue bands in NP exhibited the best performance (R2=0.798, RMSE=700.9g). Also, the models using ratio between Red edge and green bands in DP exhibited the best performance (R2=0.951, RMSE=162.4).

Discussions: It is regarded that red edge band has to be used to solve the problem because it was revealed that light-sensitive NIR band is easy to be saturated at intermediate to high weight.

Conclusion: It was possible to accurately estimate the growth of Chinese cabbage with ratio of red edge and visible bands depending on vegetation stages.